Research On Rolling Bearing Fault Diagnosis Method Based On Sparse Representation

Rolling bearing is one of the important parts in rotating machinery and equipment,and its running state will directly affect the stability of the entire equipment.Because of the structural characteristics of the rolling bearing itself and the working environment in which it occurs,faults often occur.Therefore,it is of great significance to carry out fault diagnosis research on rolling bearings.Rolling bearing fault diagnosis mainly includes signal acquisition,fault feature extraction and pattern recognition.This paper takes rolling bearing fault vibration signals as research object,and proposes a research method of rolling bearing fault diagnosis based on sparse representation.The research focus of rolling bearing fault diagnosis is to extract sensitive fault features from non-linear and non-stationary fault vibration signals.As a new signal processing technology,sparse representation can maximize the internal characteristics of the signal.Therefore,the sparse representation theory is selected to Rolling bearings for fault diagnosis studies.This article first introduces the basic principles and steps of sparse representation in detail and the ideas applied to the neighborhood of rolling bearing fault diagnosis.Aiming at the shortcomings of the sparse representation theory in the processing of rolling bearing fault data,a fault feature extraction model of local mean decomposition(LMD)and optimized orthogonal matching tracking(OMP)was proposed.The verification analysis of simulation signals and experimental data proved that This model can effectively and accurately extract fault features,which is suitable for fault diagnosis research of rolling bearings.The features extracted using sparse representation theory are generally highdimensional feature spaces composed of time and frequency domains of rolling bearing faults.In order to achieve intelligent diagnosis of rolling bearing faults,this paper uses local linear embedding(LLE)algorithm to extract high-dimensional faults extracted from sparse representation theory.Feature space is subjected to dimensionality reduction processing,combined with learning vector quantization neural network(LVQ)for fault diagnosis of rolling bearings.The experimental data are analyzed and verified.The experimental results show that the fault diagnosis model based on sparse representation and LVQ neural network can effectively perform pattern recognition on rolling bearings.Finally,the experimental data of rolling bearing is used to verify the fault diagnosis method of roller bearing based on sparse representation.The different types of rolling bearing and the damage degree of inner ring fault are diagnosed and analyzed.By comparison with BP neural network,the experimental results show that The fault diagnosis model of LVQ neural network has higher diagnostic accuracy,which proves the feasibility and practicability of sparse representation theory for fault diagnosis of rolling bearings.